# Efficient Adaptive Ensembling for Image Classification

arXiv: https://arxiv.org/abs/2206.07394 · HTML: https://mlsanity.com/paper/2206.07394/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 39 references that are in the map.
- First posted on arXiv 15 Jun 2022.

## Builds on (most cited first, 18 of 18)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Rethinking the Inception Architecture for Computer Vision](https://mlsanity.com/paper/1512.00567/) (2015)
- [Aggregated Residual Transformations for Deep Neural Networks](https://mlsanity.com/paper/1611.05431/) (2016)
- [Sharpness-Aware Minimization for Efficiently Improving Generalization](https://mlsanity.com/paper/2010.01412/) (2020)
- [Scaling Vision Transformers](https://mlsanity.com/paper/2106.04560/) (2021)
- [CvT: Introducing Convolutions to Vision Transformers](https://mlsanity.com/paper/2103.15808/) (2021)
- [Towards Federated Learning at Scale: System Design](https://mlsanity.com/paper/1902.01046/) (2019)
- [Green AI](https://mlsanity.com/paper/1907.10597/) (2019)
- [ImageNet-21K Pretraining for the Masses](https://mlsanity.com/paper/2104.10972/) (2021)
- [CoAtNet: Marrying Convolution and Attention for All Data Sizes](https://mlsanity.com/paper/2106.04803/) (2021)
- [High-Performance Large-Scale Image Recognition Without Normalization](https://mlsanity.com/paper/2102.06171/) (2021)
- [Ensemble deep learning: A review](https://mlsanity.com/paper/2104.02395/) (2021)
- [AdaBelief Optimizer: Adapting Stepsizes by the Belief in Observed Gradients](https://mlsanity.com/paper/2010.07468/) (2020)
- [Neural Architecture Transfer](https://mlsanity.com/paper/2005.05859/) (2020)
- [A Survey of Methods for Low-Power Deep Learning and Computer Vision](https://mlsanity.com/paper/2003.11066/) (2020)

## Built on by (most cited first, 8 of 8)

- [When Multi-Task Learning Meets Partial Supervision: A Computer Vision Review](https://mlsanity.com/paper/2307.14382/) (2023)
- [A logifold structure on measure space](https://mlsanity.com/paper/2405.05492/) (2024)
- [Medical Waste Sorting: a computer vision approach for assisted primary sorting](https://mlsanity.com/paper/2303.04720/) (2023)
- [Autoselection of the Ensemble of Convolutional Neural Networks with Second-Order Cone Programming](https://mlsanity.com/paper/2302.05950/) (2023)
- [Logifold: A Geometrical Foundation of Ensemble Machine Learning](https://mlsanity.com/paper/2407.16177/) (2024)
- [With Great Backbones Comes Great Adversarial Transferability](https://mlsanity.com/paper/2501.12275/) (2025)
- [ConstellationNet: Reinventing Spatial Clustering through GNNs](https://mlsanity.com/paper/2503.07643/) (2025)
- [A Critical Review on the Effectiveness and Privacy Threats of Membership Inference Attacks](https://mlsanity.com/paper/2603.22987/) (2026)

## Abstract (arXiv)

In recent times, with the exception of sporadic cases, the trend in Computer Vision is to achieve minor improvements compared to considerable increases in complexity. To reverse this trend, we propose a novel method to boost image classification performances without increasing complexity. To this end, we revisited ensembling, a powerful approach, often not used properly due to its more complex nature and the training time, so as to make it feasible through a specific design choice. First, we trained two EfficientNet-b0 end-to-end models (known to be the architecture with the best overall accuracy/complexity trade-off for image classification) on disjoint subsets of data (i.e. bagging). Then, we made an efficient adaptive ensemble by performing fine-tuning of a trainable combination layer. In this way, we were able to outperform the state-of-the-art by an average of 0.5% on the accuracy, with restrained complexity both in terms of the number of parameters (by 5-60 times), and the FLoating point Operations Per Second (FLOPS) by 10-100 times on several major benchmark datasets.